# Multimodal Question Answering on WebQA Image-grounded (test, n=2,511) Source: [Multimodal Question Answering on WebQA Image-grounded (test, n=2,511)]() Collection index: [llms.txt]() Data updated: 2026-06-11T08:05:28.299Z Evaluation of model performance on the specific test subset (n=2,511) of the WebQA Image-grounded dataset for Multimodal Question Answering, reporting F1, Accuracy, and R@k. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Multimodal Question Answering | | Dataset (including any stated split) | WebQA Image-grounded n=2,511 (test) | | Metric | F1 Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 16 of 16 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [F1 Score](): Higher is better; unit: PERCENT. - [Accuracy](): Higher is better; unit: PERCENT. - [R@k](): Higher is better; unit: PERCENT. ## Recorded results | Method | F1 Score | Result date | Report paper | | --- | --- | --- | --- | | Latent Memory | 69.4 | Not recorded | 6a28f36e90966ade919c18ff | | Latent Memory | 56.5 | Not recorded | 6a28f36e90966ade919c18ff | | Nemo-Emb-1B | 53 | Not recorded | 6a28f36e90966ade919c18ff | | Dense Retrieval | 49.8 | Not recorded | 6a28f36e90966ade919c18ff | | Qwen3-VL-Emb-8B | 34.1 | Not recorded | 6a28f36e90966ade919c18ff | | Nemo-Emb-1B | 32.9 | Not recorded | 6a28f36e90966ade919c18ff | | BM25 Retrieval | 32.6 | Not recorded | 6a28f36e90966ade919c18ff | | Latent Memory | 32 | Not recorded | 6a28f36e90966ade919c18ff | | Dense Retrieval | 30 | Not recorded | 6a28f36e90966ade919c18ff | | Qwen3-VL-Emb-8B | 20.1 | Not recorded | 6a28f36e90966ade919c18ff | | Nemo-Emb-1B | 19.9 | Not recorded | 6a28f36e90966ade919c18ff | | BM25 Retrieval | 19.2 | Not recorded | 6a28f36e90966ade919c18ff | | Dense Retrieval | 18.9 | Not recorded | 6a28f36e90966ade919c18ff | | Qwen3-VL-Emb-8B | 15.1 | Not recorded | 6a28f36e90966ade919c18ff | | BM25 Retrieval | 14.5 | Not recorded | 6a28f36e90966ade919c18ff | | Full Context | 0 | Not recorded | 6a28f36e90966ade919c18ff | ### Latent Memory F1 Score: 69.4. Result ID: 6a2a6ae0fbc373592c27534d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"5","Generator VLM":"LLaVA-1.5-13B","#Tok":"82"}. This record does not assert independent reproduction. ### Latent Memory F1 Score: 56.5. Result ID: 6a2a6ae0fbc373592c27534c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"2","Generator VLM":"LLaVA-1.5-13B","#Tok":"52"}. This record does not assert independent reproduction. ### Nemo-Emb-1B F1 Score: 53. Result ID: 6a2a6ae0fbc373592c275347. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"k":"5","Generator VLM":"LLaVA-1.5-13B","#Tok":"1885"}. This record does not assert independent reproduction. ### Dense Retrieval F1 Score: 49.8. Result ID: 6a2a6ae0fbc373592c275344. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"5","Generator VLM":"LLaVA-1.5-13B","#Tok":"1645"}. This record does not assert independent reproduction. ### Qwen3-VL-Emb-8B F1 Score: 34.1. Result ID: 6a2a6ae0fbc373592c27534a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"5","Generator VLM":"LLaVA-1.5-13B","#Tok":"957"}. This record does not assert independent reproduction. ### Nemo-Emb-1B F1 Score: 32.9. Result ID: 6a2a6ae0fbc373592c275346. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"k":"2","Generator VLM":"LLaVA-1.5-13B","#Tok":"892"}. This record does not assert independent reproduction. ### BM25 Retrieval F1 Score: 32.6. Result ID: 6a2a6ae0fbc373592c275341. [Reporting paper]() Recorded evaluation configuration: {"k":"5","Generator VLM":"LLaVA-1.5-13B","#Tok":"932"}. This record does not assert independent reproduction. ### Latent Memory F1 Score: 32. Result ID: 6a2a6ae0fbc373592c27534b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"1","Generator VLM":"LLaVA-1.5-13B","#Tok":"42"}. This record does not assert independent reproduction. ### Dense Retrieval F1 Score: 30. Result ID: 6a2a6ae0fbc373592c275343. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"2","Generator VLM":"LLaVA-1.5-13B","#Tok":"814"}. This record does not assert independent reproduction. ### Qwen3-VL-Emb-8B F1 Score: 20.1. Result ID: 6a2a6ae0fbc373592c275349. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"2","Generator VLM":"LLaVA-1.5-13B","#Tok":"465"}. This record does not assert independent reproduction. ### Nemo-Emb-1B F1 Score: 19.9. Result ID: 6a2a6ae0fbc373592c275345. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"k":"1","Generator VLM":"LLaVA-1.5-13B","#Tok":"507"}. This record does not assert independent reproduction. ### BM25 Retrieval F1 Score: 19.2. Result ID: 6a2a6ae0fbc373592c275340. [Reporting paper]() Recorded evaluation configuration: {"k":"2","Generator VLM":"LLaVA-1.5-13B","#Tok":"476"}. This record does not assert independent reproduction. ### Dense Retrieval F1 Score: 18.9. Result ID: 6a2a6ae0fbc373592c275342. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"1","Generator VLM":"LLaVA-1.5-13B","#Tok":"476"}. This record does not assert independent reproduction. ### Qwen3-VL-Emb-8B F1 Score: 15.1. Result ID: 6a2a6ae0fbc373592c275348. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"k":"1","Generator VLM":"LLaVA-1.5-13B","#Tok":"284"}. This record does not assert independent reproduction. ### BM25 Retrieval F1 Score: 14.5. Result ID: 6a2a6ae0fbc373592c27533f. [Reporting paper]() Recorded evaluation configuration: {"k":"1","Generator VLM":"LLaVA-1.5-13B","#Tok":"295"}. This record does not assert independent reproduction. ### Full Context F1 Score: 0. Result ID: 6a2a6ae0fbc373592c27533e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Generator VLM":"LLaVA-1.5-13B","#Tok":"11655"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.